Hacker News Reader

Comment by _Microft | Hacker News Reader

I found a surprising way to reduce my carbon footprint: I downloaded the music and am playing it locally now instead of streaming it over and over again.

I was listening to music on Youtube pretty much all the time and that amounted to about 35GB of traffic per month in average.

That's 420GB per year and with 7kWh/GB and emissions of approximately 500g/kWh that's a surprising total of roughly 1.5t of CO_2 emissions per year. And a downside is almost non-existant.

Edit: my playlist consists of a small number of songs, so I actually get savings. That would not work when listening to a list that does not actually repeat of course.

View on HN

There is no way 7kWh/GB is an accurate figure. For example, AWS Oregon region charges you $0.05/GB after the first 150 TB.[1] This page[2] says the current electricity rate of Oregon is 9.8¢/kWh. Assuming AWS is getting electricity at half the price (and making no profit), it can burn 1.02 kWh/GB.

The actual number is most likely lower: AWS has to make some profit, and it's not known for being cheap.

[1] https://aws.amazon.com/s3/pricing/ [2] https://www.electricitylocal.com/states/oregon/

The article talks about the cost of transferring a file from the cloud vs loading it from a local harddrive.

It's not only the amount of energy used for that at the datacenter, it's also summing up all the fractions of power usage per GB of the used fiber connection, the repeaters on this fiber connection, routers, switches, losses in power supplies, energy required for air conditioning the hardware, the phone/cable/mobile infrastructure that also has switches, repeaters, ... until ones own modem, router and computer at the own location that is needing energy to operate.

It really does not matter how much they pay because the market price does not include the actual costs a company imposes on the environment. There is no CO2 Tax, is there?

Ask yourself: How can it be more efficient to stream the same stuff over and over again, wasting CPU cycles on couple of fat machines etc. instead of playing it from a local cache (hard drive, ...). IMHO it cannot be more efficient. It is convenient and it appears to be cheap.

You are right that these companies don't pay the "actual costs" on environmental damage, but that's a separate issue, and not really relevant to the initial claim. If you claim that someone is using 7 kWh or electricity, you had better be prepared to show they actually do. (Besides, consumers don't pay for the environmental cost of building and shipping an SSD from Taiwan, either.)

Companies like Amazon may not care about environment, but they care a lot about saving money, and modern computers can handle a surprising amount of connections simultaneously: it's not like those beefy AWS instances are serving just your music.

The 7kWh figure was not mine. However, I see what the intend of the Thread starter was and I think he is right. Streaming video/music on top of 'AI' (to sell more stuff), security theater and Ad campaigns is just a waste of resources.

/s I would bet Bezos and folks can buy fresh air when it comes to this point. But at least we all could watch Game of Thrones when we wanted to.

Where on earth did you dig up the 7kWh/GB figure from?

It was mentioned in [0], it's actually the higher one of two different sources (I haven't looked at the sources to be honest because that would require actually finding them which I haven't yet):

A Carnegie Mellon University study concluded that the energy cost of data transfer and storage is about 7 kWh per gigabyte. An assessment at a conference of the American Council for an Energy-Efficient Economy reached a lower number: 3.1 kWh per gigabyte.

[0] https://stanfordmag.org/contents/carbon-and-the-cloud (2017)

Those numbers are way off. The article cites a study claiming 5 kWh here:

https://aceee.org/files/proceedings/2012/data/papers/0193-00...

It's from 2012 and cites numbers from as long ago as 2009 to determine normalized energy intensity. It also combines different numbers in a way that mean you can't estimate how much energy is saved by downloading a gigabyte to local storage instead of repeatedly streaming it. The single largest energy segment in their estimate is actually desktop computers! And it counts the energy used by CPU-intensive scripts along with the energy required by actual data transfer.

Sanity check: Cisco estimates "busy hour" internet traffic at around 3 petabits per second in 2019.

https://www.cisco.com/c/en/us/solutions/collateral/service-p...

If it took 5 kWh/GB for data transfer during that busy hour, corresponding peak hour electricity use would be in excess of 6.5 terawatts. But that's more electricity than the world can generate in any given hour.

Even if the numbers are way off it is obvious that watching TV or listening music over a packet based protocol is way more energy consuming than any UKW sender can be. UKW scales - streaming does not without more energy because you need more computers.

UKW?

Ultra Kurzwelle (ultra short radio wave). No idea what the translation to English is.

It's an absurd number that can't really be considered as is.

Does that include storage for 100 years? Or just the few seconds needed for transfer.

Do they count all of the datacenters energy usage as something needed to just transfer and store the data (with no other processing)?

Unfortunately that article didn't answer anything.

I'm surprised to hear it takes 7 kWh to transmit 1 GB. My phone consumes only 0.3 kWh a month and I use 1 GB. So the power requirements are very asymmetric.

That probably factors in the infrastructure it takes to deliver the data.

Yes, it is comparing what it takes to load a file from disk or from remote. It needs to include the share of the used infrastructure for that.

Reply on news.ycombinator.com